some thoughts on the Modular Project

From Haible
Sent Mon, Mar 20th 1995, 21:51

Well, I've been listening about this for a while ...
if You like some comments, here they are.

It's a geat thing You're obviously going to start. I think You
know that the greatest problems that will occur in a project
like this will *not* be of electronic nature. Anyway, it's the
electronics I'd like to make some suggestions about.

> Does anyone have a preference as far as the control voltage standard?
> The human ear can (when young and perfect) hear about 10 octaves.  If
> memory serves me correctly some older synths used -5 volts to +5 volts
> to represent this and some used 0 volts to 10 volts.

This is less a problem than You think, IMO. If You think about a 5-octave
keyboard, You'd have a 0 ... 5V range. This is compatible with all Midi/CV
converters I know as well. Any additional octave shifting usually is done
with Transpose or Coarse Frequency switches/knobs, and by additional
modulation CV inputs. *Internally*, the 1V/Oct standard has to be changed
to a 19mV/Oct, anyway. Just make sure that non of Your internal summing
stages run into the supply rails for the whole frequency range. (I made that
mistake with my first throw of the Phaser/Filter module!)
With 10V for audio range and +/-15V supplies there should be enough
"headroom" even for sub-audio (LFO etc.) applications with just a little 
in the design.

> Also, what should the standard for gates be?  TTL (standard computer
> chip) levels with low = open?  Many older synths had gate pulses be 12
> or 15 volts with high = open.  Should clocked modules (like
> sequencers) clock on the rising or falling edge?  If gates are low =
> open then it should probably be falling edge so that you can use the
> gate to manually clock it.  These issues are what makes doing modular
> work with equipment from multiple manufacturers a nightmare.  We
> should figure all of this out before getting started on circuit
> design.

I highly recommend not to use special gate signals or digital IC families
*at all*. This might sound a little bold, but take it that way: What should
be the difference between (1) a gate signal, (2) a clock signal, (3) a LFO
pulse waveform or even (4) a VCO pulse wave ?? In the real world,
unfortunately, they often are different. But they needn't be, If You want
it. Personally, I do it like that:
My system's signal for "gating" (in the widest sense) is 5V (on) 0V (off),
with positive edge for triggers. This is the most common standard, and
if I have to include other standards, I have to add some little convertors,
anyway (more on this, see below). Everything else depends on that
decision: I choose the pulse wave outputs of my VCO's to be 0V/5V.
The same with LFO's. so everything *inside* is compatible at least.
Now I want to describe my input and output stages:
All gate inputs are schmitt-triggers that turn on at about 2V and turn off
below about 1V or 1.5V. This costs me 2 extra transistors in every
gate input, but it makes it compatible to almost *everything* in the
world, besides switch trigger (of course). If I had logic IC's instead, I'd
have to fear for overload (with TTL and 15V inputs ...) or for too high
thresholds (15V-CMOS won't accept TTL signals). The best thing of
the schmitt-trigger approach, however, is that you can feed the input
with other than pulse waves. Feed it with a LFO's sawtooth and it will
work. Change the amplitude of the saw, and You have control of the
pulse widhth or gate duration. Route the saw wave over a VCA into
the input, and You have voltage controlled gate duration. Feed it with
a triangle, summed with a certain amount of DC, You have a trigger
delay for free. Route the DC offset thru a VCA and You have ...
You see what I mean.
The output stages are a little more tricky. In my first modular, I used
a seperate 5V supply for the gates. I wouldn't do that again. If You
manage to get a scaled-down saw or triangle waveform in a VCO,
why should the same thing be a problem with pulses and for gates ??
It is important, that a gate output can both, sink and source, some
milliamperes, to interface with any type of 5V gate inputs. So the best
thing to get this from a +/-15V power supply is using 15V-locic
internally (40xx CMOS stuff, which also has advantages in driving
CMOS electronic switches...), scaling it down with a rather high-
impedance voltage divider (10kOhm range ...) and then buffering it
with an ordinary OpAmp. Don't forget to add a small series resistor
(100 ... 500 OHM) at the OpAmp output to prevent oscillation due
to capacitive load.
The great advantages of this design are:
* No need for extra 5V supply
* capability to source and sink a decent load makes this usable
  together with S-trigger Ins, if You just do a logic inversion.
* If You have a piece of gear that really *needs* more than 5Volts
  at its gate input, just change a single resistor (in the high-Z
  voltage divider).

In short, a typical "gate type" module of mine works like that:

input -> schmitt-trigger -> internal processing with 15V CMOS-
compatible logic -> high-Z attenuator -> Opamp(Voltage follower)
 -> output.

This approach works well for me; it's a *little* bit of overhead, but
it's worth it all if You have to (or want to) interface some exotic
And the AH Modular Synthesizer won't go the "Doepfer minimum
number of components" way, will it ?!

These were my first 2 cents; others to come, perhaps.
Any comments, ideas, flames, welcome.

And GOOD LOOK for Your Modular Project!